Multi-style image generation method based on feature fusion

Abstract

The invention discloses a multi-style image generation method based on feature fusion, and belongs to the field of computer vision. The implementation method comprises the following steps: inputting a semantic segmentation graph into a content feature extraction network, and extracting a content feature vector in the semantic graph; inputting the style graph into a style feature extraction network, and extracting a style feature vector in the style graph; inputting the extracted content feature vector fc and the extracted style feature vector fs into a content style feature fusion network for feature fusion to obtain a fusion feature vector after feature fusion; constructing a generative adversarial network composed of a generator and a discriminator, and training the generative adversarial network on the data set by designing a loss function; and generating a multi-style image with the semantic graph content and the style of the style graph by using a trained generator with a minimized loss function. According to the invention, the generated multi-style image with the semantic graph content and the style of the style graph can be applied to a scene attracting attention, and related engineering technical problems are solved.

Description

technical field [0001] The invention relates to an image generation method for generating a multi-style image from a semantic segmentation map, in particular to an end-to-end rapid generation method capable of realizing a semantic map to a multi-style image, belonging to the field of computer vision. Background technique [0002] At present, most of the models that generate multi-style images are generated from real images, but the few models that generate style images from semantic maps can only use images in the same data set as input styles, and cannot achieve the same style. Quick migration of any style. [0003] End-to-end generation of images of any style from the semantic map is of great significance in the direction of art design and virtual reality education resource generation. In the field of art design, art creators or designers can only specify the position of each object in the semantic map And the general shape and the style you want to generate, then you can...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problem that the generation of multi-style images from semantic graphs has great limitations in the background technology, the technical problem to be solved by a multi-style image generation method based on feature fusion disclosed in the present invention is: to provide a network with content feature extraction, style features The network framework for generating style images from semantic graphs is composed of three parts: the extraction network and the content style feature fusion network. The content features and style features are extracted respectively through the content feature extraction network and the style feature extraction network. The features extracted by the two networks are fused to generate a multi-style image with semantic map content and style map style

Images